Article Information

• PDF (800 kb)
• Full Text with Large Figures
• Supporting Material
• Cited by in Scopus (4)

PubMed

• Articles by Gilad Gibor
• Articles by Bernard Attali

Related Articles

• Determinants within the Turret and Pore-Loop Domains of KCNQ...

  Determinants within the Turret and Pore-Loop Domains of KCNQ3K Channels Governing Functional Activity Biophysical Journal, Volume 95, Issue 11, 1 December 2008, Pages 5121-5137 Oleg Zaika, Ciria C. Hernandez, Manjot Bal, Gleb P. Tolstykh and Mark S. Shapiro Abstract KCNQ1–5 (Kv7.1–7.5) subunits assemble to form a variety of functional K channels in the nervous system, heart, and epithelia. KCNQ1 and KCNQ4 homomers and KCNQ2/3 heteromers yield large currents, whereas KCNQ2 and KCNQ3 homomers yield small currents. Since the unitary conductance of KCNQ3 is five- to 10-fold greater than that of KCNQ4 or KCNQ1, these differences are even more striking. To test for differential membrane protein expression, we performed biotinylation and total internal reflection fluorescence imaging assays; however, both revealed only small differences among the channels, leading us to investigate other mechanisms at work. We probed the molecular determinants governing macroscopic current amplitudes, with focus on the turret and pore-loop domains of KCNQ1 and KCNQ3. Elimination of the putative N289 glycosylation site in KCNQ1 reduced current density by ∼56%. A chimera consisting of KCNQ3 with the turret domain (TD) of KCNQ1 increased current density by about threefold. Replacement of the proximal half of the TD in KCNQ3 with that of KCNQ1 increased current density by fivefold. A triple chimera containing the TD of KCNQ1 and the carboxy terminus of KCNQ4 yielded current density 10- or sixfold larger than wild-type KCNQ3 or KCNQ1, respectively, suggesting that the effects on current amplitudes of the TD and the carboxy-terminus are additive. Critical was the role of the intracellular TEA-binding site. The KCNQ3 (A315T) swap increased current density by 10-fold, and the converse KCNQ1 (T311A) swap reduced it by 10-fold. KCNQ3 (A315S) also yielded greatly increased current amplitudes, whereas currents from mutant A315V channels were very small. The KCNQ3 (A315T) mutation increased the sensitivity of the channels to external Ba block by eight- to 28-fold, consistent with this mutation altering the structure of the selectivity filter. To investigate a structural hypothesis for the effects of these mutations, we performed homology modeling of the pore region of wild-type and mutant KCNQ3 channels, using KvAP as a template. The modeling suggests a critical stabilizing interaction between the pore helix and the selectivity filter that is absent in wild-type KCNQ3 and the A315V mutant, but present in the A315T and A315S mutants. We conclude that KCNQ3 homomers are well expressed at the plasma membrane, but that most wild-type channels are functionally silent, with rearrangements of the pore-loop architecture induced by the presence of a hydroxyl-containing residue at the 315 position “unlocking” the channels into a conductive conformation. Abstract | Full Text | PDF (1438 kb)

• Voltage-gated K Channels - IV

  Voltage-gated K Channels - IV Biophysical Journal, Volume 94, Issue , 1 February 2008, Pages 459-467 Full Text | PDF (115 kb)

• Differential Roles of S6 Domain Hinges in the Gating of KCNQ...

  Differential Roles of S6 Domain Hinges in the Gating of KCNQ Potassium Channels Biophysical Journal, Volume 90, Issue 6, 15 March 2006, Pages 2235-2244 Guiscard Seebohm, Nathalie Strutz-Seebohm, Oana N. Ureche, Ravshan Baltaev, Angelika Lampert, Ganna Kornichuk, Kaichiro Kamiya, Thomas V. Wuttke, Holger Lerche, Michael C. Sanguinetti and Florian Lang Abstract Voltage-gated K channel activation is proposed to result from simultaneous bending of all S6 segments away from the central axis, enlarging the aperture of the pore sufficiently to permit diffusion of K into the water-filled central cavity. The hinge position for the bending motion of each S6 segment is proposed to be a Gly residue and/or a Pro-Val-Pro motif in Kv1–Kv4 channels. The KCNQ1 (Kv7.1) channel has Ala-336 in the Gly-hinge position and Pro-Ala-Gly. Here we show that mutation of Ala-336 to Gly in KCNQ1 increased current amplitude and shifted the voltage dependence of activation to more negative potentials, consistent with facilitation of hinge activity that favors the open state. In contrast, mutation of Ala-336 to Cys or Thr shifted the voltage dependence of activation to more positive potentials and reduced current amplitude. Mutation of the putative Gly hinge to Ala in KCNQ2 (Kv7.2) abolished channel function. Mutation-dependent changes in current amplitude, but not kinetics, were found in heteromeric KCNQ1/KCNE1 channels. Mutation of the Pro or Gly of the Pro-Ala-Gly motif to Ala abolished KCNQ1 function and introduction of Gly in front of the Ala-mutations partially recovered channel function, suggesting that flexibility at the PAG is important for channel activation. Abstract | Full Text | PDF (437 kb)

• …more

Copyright © 2007 The Biophysical Society. All rights reserved.
Biophysical Journal, Volume 93, Issue 12, 4159-4172, 15 December 2007

doi:10.1529/biophysj.107.107987

Channels, Receptors, and Electrical Signaling

An Inactivation Gate in the Selectivity Filter of KCNQ1 Potassium Channels

Gilad Gibor^*, 1, Daniel Yakubovich^*, 1, Avia Rosenhouse-Dantsker^†, 1, Asher Peretz^*, Hella Schottelndreier^*, Guiscard Seebohm^‡, Nathan Dascal^*, Diomedes E. Logothetis^†, Yoav Paas^§, ,   and Bernard Attali^*, ,  

^* Department of Physiology and Pharmacology, Sackler Medical School, Tel Aviv University, Tel Aviv, Israel
^† Department of Physiology and Biophysics, Mount Sinai School of Medicine, New York, New York
^‡ Physiologisches Institut I, Universität Tübingen, Tübingen, Germany
^§ The Mina and Everard Goodman Faculty of Life Sciences Bar-Ilan University, Ramat-Gan, Israel

Address reprint requests to Bernard Attali, PhD, Tel.: 972-3640-5116 or Yoav Paas, PhD, Tel.: 973-3531-7968.

¹ Gilad Gibor, Daniel Yakubovich, and Avia Rosenhouse-Dantsker contributed equally to this work.

Potassium channels catalyze fast transfer of K⁺ ions across the cell membrane at nearly diffusion-limit rates while selecting K⁺ over Na⁺ by >100-fold. Structure determination at atomic resolution of bacterial and more recently mammalian K⁺ channels has proven to be a landmark achievement in the field of ion channels 1,2. K⁺ channels comprise four subunits arranged symmetrically around an ion-conducting pore. In voltage-gated K⁺ channels (Kv), each subunit consists of six transmembrane segments, including an S5–S6 region encompassing the aqueous pore and a peripheral S1–S4vol.age sensor domain 3. The narrowest part of the pore is the selectivity filter which achieves a remarkably fast and selective permeation of K⁺ ions. The filter’s geometry, a tunnel-like structure of 12Å length and 3Å diameter, provides an efficient permeation pathway for dehydration, diffusion, and rehydration of K⁺ ions 4. To this end, the selectivity filter of K⁺ channels is endowed with four K⁺ coordination sites. The first three sites s₁–s₃ are formed by four rings of backbone carbonyl oxygen atoms, while the inner s₄ site is formed by both backbone carbonyl oxygens and side-chain hydroxyl oxygens 5,6. The K⁺ coordination geometry in the selectivity filter mimics the coordination of K⁺ ions by water molecules. Under physiological K⁺ concentrations, two K⁺ ions reside in the selectivity filter at a given time and move in a highly concerted fashion between two isoenergetic configurations 7.

Several lines of evidence suggest that the selectivity filter has also the ability to act as a gate by switching between conducting and nonconducting conformations 8,9,10,11,12. Most recent studies revealed that the selectivity filter of the bacterial KcsA K⁺ channel can undergo large conformational excursions associated with transitions between closed, open, and inactivated states 13,14.

Inactivation of Kv channels can occur by N-type or/and C-type mechanisms. The fast N-type inactivation involves an intracellular peptide domain at the N-terminus of α- or β-subunits which occludes the channel central cavity and prevents ion permeation 15,16,17. The slow C-type inactivation was suggested to involve structural rearrangements in the outer pore leading to a loss of K⁺ coordination sites in the selectivity filter 18,19,20,21. The biophysical hallmarks of C-type inactivation are reflected by its inhibition by high external K⁺ or external tetraethylammonium 15,22,23. These features have been interpreted as a foot-in-the-door mechanism, in which occupancy of an ion binding site by K⁺ or TEA at the external filter entrance slows or prevents the conformational changes required for C-type inactivation 19. It should be noted that in Shaker channels, a Ba²⁺ site was identified topologically below the C-type inactivation gate 24.

KCNQ1 channels (Kv7.1) belong to a subfamily of voltage-gated K⁺ channels, Kv7, and coassemble with KCNE1 β-subunits to generate the I_KS potassium current that is critical for normal repolarization of the cardiac action potential 25,26,27,28. Mutations in either KCNQ1 or KCNE1 genes produce the long QT syndrome (LQT), a human ventricular arrhythmia 26,27,29. Inactivation of KCNQ1 channels does not exhibit the hallmarks of N- and C-type inactivation. Inactivation of wild-type (WT) KCNQ1 channels is invisible macroscopically but can be revealed by hooked tail currents which reflect recovery from an inactivation process 30,31,32. Here, we characterized KCNQ1 pore mutants, including the LQT mutation L273F, which exhibit a voltage-dependent slow inactivation. We found that this slow inactivation delays entry of Ba²⁺ ions to the pore and trap them by slowing their exit from the selectivity filter. External potassium ions accelerate inactivation kinetics and exacerbate barium trapping. Our experimental data together with molecular modeling suggest that the voltage-dependent slow inactivation arises from a reduced flexibility of the carbonyl ring of Tyr³¹⁵ and a stronger coordination of K⁺ at filter site s₁.

Female Xenopus Laevis frogs were purchased from Xenopus 1 (Dexter, MI). The procedures followed for surgery and maintenance of frogs were approved by the animal research ethics committee of Tel Aviv University and in accordance with the Guide for the Care and Use of Laboratory Animals (1996, National Academy of Sciences, Washington, DC). Frogs were anaesthetized with 0.15% tricaine (Sigma, St. Louis, MO). Pieces of the ovary were surgically removed and digested with 1mg/ml collagenase (type IA, Sigma) in Ca²⁺-free ND96 for ∼1h, to remove follicular cells. Stage V and VI oocytes were selected for cRNA or DNA injection and maintained at 18°C in ND96 (96mM NaCl, 2mM KCl, 1.8mM CaCl₂, 1mM MgCl₂ and 5mM HEPES titrated to pH=7.5 with NaOH), supplemented with 1mM pyruvate and 50μg/ml gentamycin. The human KCNQ1 cDNA (in pGEM vector) was linearized by Not1. Capped complementary RNA was transcribed by the T7 RNA polymerase, using the mMessage mMachine transcription kit (Ambion, Austin, TX). The cRNA size and integrity was confirmed by formaldehyde-agarose gel electrophoresis. Homomeric expression of human KCNQ1 was performed by injecting 40nl per oocyte (5ng cRNA) using a Nanoject injector (Drummond Scientific, Broomall, PA). Several expression experiments were also carried out by microinjecting a recombinant DNA vector (pcDNA3) encoding the human KCNQ1 cDNA directly into Xenopus oocyte nuclei (1ng into10nl).

Standard two-electrode voltage-clamp measurements were performed at room temperature (22–24°C) 2–5 days after cRNA or DNA microinjection. Oocytes were placed into a 100μl recording chamber and superfused with a modified ND96 solution (containing 0.1mM CaCl₂) using a fast perfusion system which operates under controlled N₂ pressure allowing constant perfusion velocity of 3.9–4.2ml/min. The exchange of solutions was performed by computer-controlled pinch valves (ALA-VM8; ALA Scientific Instruments, Westbury, NY). A homemade manifold having virtually no void volume and very narrow connecting tubes prevented backward flow upon valve switch. The bath solution was completely replaced within 1.5s, allowing a solution exchange time of ∼25ms around the oocyte. Whole-cell currents were recorded using a GeneClamp 500 amplifier (Axon Instruments, Foster City, CA). Stimulation of the preparation and data acquisition were performed using the pCLAMP 6.02 software (Axon Instruments) and a 586 personal computer interfaced with a Digidata 1200 interface (Axon Instruments). Glass microelectrodes (A-M Systems, Carlsborg, WA) were filled with 3M KCl and had tip resistances of 0.2–0.5MΩ. Current signals were digitized at 1kHz and low-pass-filtered at 0.2kHz. The holding potential was −80mV. Leak subtraction was performed off-line, using steps from −120 to −90mV, assuming that the channels are closed at −80mV and below. Errors introduced by the series resistance of the oocytes were not corrected and were minimized by keeping expression of the currents below 10μA. All BaCl₂ solutions were prepared in modified ND96 (containing 0.1mM CaCl₂) and were isoosmotically changed for NaCl. Modified ND96 solutions containing high K⁺ concentrations (50mM) were also isoosmotically changed for NaCl.

Data analysis was performed using the Clampfit program (pCLAMP 8, Axon Instruments), Microsoft Excel 2002 (Microsoft, Redmond, WA), SigmaPlot 8.0 (SPSS, New York, NY) and Prism 4 (GraphPad Software, San Diego, CA). To analyze the voltage dependence of channel activation, a double exponential fit was applied to the tail currents at −60mV or −120mV and the slow exponential component was extrapolated to the beginning of the repolarizing step. Chord conductance (G) was calculated by using the equation

(1)

(2)

(3)

The voltage-dependence of Ba²⁺ block was calculated according to the Woodhull’s model 33 and the voltage dependence of the dissociation constant is given by

(4)

Initially, we calculated the conductance time course by dividing the current values observed during voltage step from −80mV to +20mV by the driving force assuming that the reversal potential for K⁺ in Xenopus oocytes is −98mV.

The time course of activation of WT KCNQ1 could be described as a Markov process 31,34,35 by Scheme I, in which k values are the rate constants in s⁻¹ and C₁ to I₁ are the corresponding state occupancies. Only O₁ and O₂ are assumed to conduct. Scheme I is sufficient for the description of macroscopic behavior of KCNQ1 channels 31, assuming that the stochastic behavior of a single channel is reflected in the macroscopic behavior of a channel population 36.

Display large version of this figure

Scheme I

k-values are the rate constants in s⁻¹ and C₁ to I₁ are the corresponding state occupancies. Only O₁ and O₂ are assumed to conduct.

Display large version of this figure

Scheme II

Characterized by two inactivated states: 1), fast inactivated state I₁, similar to that of the wild-type; and 2), slow inactivated state I₂, which can be accessed from two different open states O₂ and O₁, respectively.

The time course of activation of L273F mutant was characterized by two inactivation processes: 1), fast inactivation, similar to that of the wild-type; and 2), slow inactivation. Among several possible schemes of activation we have chosen Scheme II, which includes both inactivation processes and assumes that the inactivation states can be accessed from different open states (see Results).

We generated a set of differential equations based on Scheme II and fitted it to the observed time-course of the L273F mutant activation (to obtain the transition rate constants and relative occupancies of each state). For simplicity, we assumed that no openings occur when the cell is voltage-clamped to −80mV (the holding potential) and consequently we fixed initial values of all states to zero, except C₁.

According to Scheme II, the open channel probability can be calculated from

(5)

(6)

For the simulation, we used the averaged values of the rate constants and the relative state occupancies obtained from the fit of the activation time course during the +20-mV voltage step and allowed the simulation to run for predefined time intervals. Thereafter, we used the simulated values and changed the rate constants to those corresponding to the −60-mV voltage step.

Time-course fitting and simulation were done with Berkeley Madonna software (Ver. 8.0.2 for Windows; Kagi Shareware, Perth, Western Australia).

As a first step, the sequence of the S5-pore-S6 segment of the human KCNQ1 channel (Ile²⁵⁷–Gln³⁵⁷; SWISS-PROT entry P51787) was submitted for searching a homologous template in the SWISS-MODEL repository (http://swissmodel.expasy.org/SWISS-MODEL.html). The search for homologous sequences of known three-dimensional structure scored the mammalian Shaker Kv1.2 potassium channel (PDB ID code 2A79) with the highest probability to match as a structural template. The KCNQ1 sequence was therefore aligned with the Ala³²³–Arg⁴¹⁹ sequence of the Shaker Kv1.2 channel using the program T-COFFEE 37 (Supplementary Material Fig. 3 ), and the alignment was submitted to automated comparative protein modeling via the SWISS-PROT alignment interface 38. The root mean-square difference between the KCNQ1 structural model and the template (Shaker Kv1.2) was 0.11Å for 96Cα atoms of the aligned amino acids (0.27Å for 384 backbone atoms). Note that this superposition does not include amino acids 290–294, which form an extra loop structure owing to an alignment gap in the turret region. Four identical subunit models were organized around the axis of K⁺ conduction by three-dimensional superposition of 188 backbone atoms of amino acids Thr³¹¹–Gln³⁵⁷ onto the homologous segment, Thr³⁷³–Arg⁴¹⁹, of the Shaker Kv1.2 x-ray crystal structure. Energy minimization of the tetrameric KCNQ1 model was performed in vacuum with GROMOS96. No clashes within the individual subunit or at the subunit interfaces have been observed before or after energy minimization. The models of KCNQ1 mutants were elaborated by introducing the specific mutation in all subunits of the tetrameric KCNQ1 structural model before subjecting the models to energy minimization. In the case of mutant L273F, the side chain of Phe was oriented in the same direction as the side chain of the replaced Leu residue. For Trp substituted at position 273, only one side-chain conformation was energetically favored, as all other side-chain rotamers drastically clashed with the pore helix, even after a step of energy minimization. For the models of the other mutants, see Supplementary Material .

The primary hydration shell method 39 is a method that provides both an efficient representation of solvation effects and a flexible nonspherical restraining potential. To replace the bulk representation of the solvent, a restraining force that balances the instantaneous pressure inside the primary solvent shell is applied. The molecular dynamics (MD) simulations were performed using CHARMM, version 26 40. We used structurally restrained MD simulations that maintain the structural integrity of the helical TM domain. Specifically, we imposed NOE type restraints on α-helical backbone distances as well as harmonic restraints on the Cα atoms of residues in the outer helices. The region we examined was comprised of the extracellular loops and the selectivity filter. This region was surrounded by a 30Å sphere of water molecules whose origin was located at its center of mass. Water molecules located within 2.6Å or further than 15Å from these residues were deleted and the combined system was subjected to energy minimization. Minimizations were performed using the steepest-descent and the adopted-basis Newton Raphson algorithms. After removing all waters located 5.5Å from the examined region, the system was further minimized. Three K⁺ ions were then placed at sites s₀, s₂, and s₄ of the selectivity filter, and the system was minimized again. The structure was then equilibrated and an MD simulation was carried out for 1.5ns with a time-step of 0.5 fs, and with a restraining force of 0.95Kcal/moleÅ. The analysis was done using structures recorded every 0.5ps with Origin6.1. Statistical analysis of the third and fourth moments was performed using JPM6.0.0 (statistical discovery, SAS Institute, Cary, NC), which is based on the definition of Fisher’s skewness and kurtosis 41.

Inactivation of WT KCNQ1 is invisible (hidden) in macroscopic inspection but the presence of hooked tail currents suggests a recovery from an inactivation process 30,31,32,35. However, several KCNQ1 pore mutants also display a voltage-dependent slow inactivation 42,43,44,45. Here we examined the inactivation properties of the KCNQ1 long QT mutation L273F residing in the S5 segment 45. In both WT and mutant L273F channels, the tail currents exhibit ascending and descending phases which reflect the fast recovery from a fast inactivation process and channel deactivation, respectively (Figure 1AB, and Figure 2AC). In addition, the L273F mutant generates a voltage-dependent slow inactivation which develops at depolarized potentials above −10mV (Figure 2B), with no change in K⁺ selectivity (at 2mM external K⁺, mutant L273F E_rev=−91±4mV; n=6, and WT E_rev=−93±5mV; n=8). This raises the question of whether the fast and slow inactivation of the L273F mutant originate from the same process. To address this issue, careful kinetic analyses of the relaxation and recovery from inactivation were performed.

Display large version of this figure

Figure 1

Two distinct inactivation states coexist in the KCNQ1 mutant L273F. (A) Representative trace from a triple pulse protocol reveals the fast hidden inactivation produced by WT KCNQ1 (dashed inset). (B) Representative trace of slow inactivation of L273F at +30mV, with hooked tail currents at −60mV. (C) Single-exponential time constants of fast inactivation for WT and L273F channels (left) measured as in panel A (+30mV), and of slow inactivation of L273F (right) (+30mV); note the different scales. (D) Representative trace of a twin pulse protocol measuring recovery from slow inactivation of L273F. (E) Recovery from slow inactivation of L273F as a function of time at −60mV. (F) Time constants of recovery from fast inactivation of WT and L273F measured from hooked tail currents at −60mV (left) and from slow inactivation of L273F (right), measured as in panel C. Data shown are mean±SE (n=6–8).

Display large version of this figure

Figure 2

Discrimination between fast and slow inactivations by barium. (A and B) Representative I-V traces of WT (A) and L273F (B) in the absence (control) and presence of 10mM Ba²⁺. Currents were evoked by depolarizing steps from −70mV to +30mV in 10mV increments (holding potential, −80mV; tail potentials −60mV). (C) Normalized tail currents of WT KCNQ1 obtained at −60mV and −120mV in the absence and presence of Ba²⁺ after a +30mV prepulse (n=8). (D) Simulation of current traces of WT and mutant L273F, according to gating Scheme I and Scheme II, respectively. (E) Simulation of recovery from slow inactivation of mutant L273F according to Scheme II.

The relaxation time of fast inactivation could be revealed by a triple pulse protocol (Figure 1A): a first depolarizing pulse to +30mV opens the channels of which a yet undetected portion inactivates; then, a brief (15ms) hyperpolarizing interpulse at −130mV enables recovery from channel inactivation; a third depolarizing pulse at +30mV reopens and reinactivates the channel population that previously underwent inactivation, leading to an initial extra-current. The fast relaxation measured at the third pulse corresponds to the kinetics of fast inactivation with single-exponential time constants of τ=18±1ms and τ=57±5ms for WT and mutant L273F, respectively (Figure 1A, inset, and C, left; n=6–8). In contrast, the relaxation time course of slow inactivation in mutant L273F was much slower with τ=2554±264ms (Figure 1BC, right; n=6; p<0.001).

The distinction between the two types of inactivation was also revealed by the kinetics of recovery from channel inactivation. Recovery from fast inactivation was a relatively fast process with τ=40±4ms and τ=113±9ms for WT and L273F, respectively (at −60mV) (Figure 1BF, left; n=6–8). In contrast, recovery from slow inactivation of mutant L273F was a much slower process with a time constant τ=3340±210ms, as measured by a twin pulse protocol at −60mV (Figure 1D–F; n=6; p<0.001). These data indicate that two distinct inactivation states (fast and slow) coexist in mutant L273F.

The effect of external Ba²⁺ ions on pore properties allowed us to further discriminate between the two types of inactivation. We previously showed that steady-state Ba²⁺ binding prevents the fast inactivation of WT and L273F (this study) by suppressing both the hook of the tail currents (Figure 2C) and the fast relaxation measured by the triple pulse protocol 34. In contrast, steady-state Ba²⁺ binding barely affects the extent of slow inactivation elicited by mutant L273F which displays, at +30mV, a fractional slow inactivation of 0.65±0.05 and 0.55±0.05 in the absence and presence of 10mM Ba²⁺, respectively (Figure 2B; n=6). Notably, we observed similar affinity and fractional electrical distance of Ba²⁺ block in WT and L273F channels (Kd_−40mV=0.37±0.08mM; δ=0.33 and Kd_−40mV=0.33±0.06mM; δ=0.33, respectively, n=8) (Fig. 3). The latter results suggest that the binding of Ba²⁺ deep in the pore (filter site s₄) 46 is similar in the inactivating L273F mutant and the WT.

Display large version of this figure

Figure 3

Steady-state Ba²⁺ block of mutant L273F channels. (A) Current-voltage relations of mutant L273F in the absence (squares), and presence of Ba²⁺ at 2mM (up-triangles), 10mM (down-triangles), and 20mM (diamonds). (B) Ba²⁺ inhibition curves of fractional channel unblock plotted as a function of Ba²⁺ concentrations, measured at −40mV (squares), −30mV (up-triangles), and −20mV (down-triangles). K_D values were obtained from the fit of a sigmoidal dose-response function: K_D−40mV=0.33±0.06mM; K_D−30mV=0.35±0.04mM; and K_D−20mV=0.47±0.06mM. (C) Ba²⁺ inhibition curves expressed as in panel B and measured at −10mV (open up-triangles), 0mV (open diamonds), +10mV (open down-diamonds), and +20mV (solid diamonds). K_D−10mV=0.63±0.07mM; K_{D 0mV}=0.87±0.09mM; K_D+10mV=1.28±0.12mM; and K_D+20mV=1.52±0.16mM. (D) K_D values obtained from a single-site fit of the Ba²⁺ inhibition curves plotted as a function of membrane voltage and yielding δ=0.33. Data shown are mean±SE (n=8).

Unlike the previously described gating behavior of WT KCNQ1, which accounts for only one inactivation state as shown in Scheme I31,34,35, we elaborated a kinetic model that accounts for two distinct inactivation states in mutant L273F, as shown in Scheme II. The latter assumes two closed states (C₁ and C₂), two voltage-dependent open states (O₁ and O₂), a fast voltage-independent inactivated state I₁ (fast inactivation) and a slow voltage-dependent inactivated state I₂ (slow inactivation) (Supplementary Material Table 1 ). Simulating the gating behavior of mutant L273F according to Scheme II, generates current kinetics that closely matches the experimental traces with coexistence of slow and fast inactivation (Figure 2D). Consistently, Scheme II closely simulates the experimental data measuring the kinetics of recovery from slow inactivation (Figure 1DE and Figure 2DE).

As native cardiac I_KS-channels result from the coassembly of KCNQ1 and KCNE1 subunits, we coexpressed the mutant L273F with KCNE1. Notably, KCNE1 could not completely abolish the marked inactivation of mutant I_KS-channels which leads to a significant decrease of the current amplitude (56% inhibition, p<0.01; Supplementary Material Fig. 1 ).

Ba²⁺ and K⁺ have similar ionic radii (1.35Å and 1.33Å, respectively), a physical attribute extensively used to dissect K⁺ channel permeation. To probe the pore properties of KCNQ1 mutants displaying voltage-dependent slow inactivation, we used a fast perfusion system 34,47 and monitored the kinetics of Ba²⁺ block and unblock after an activation step that opens and inactivates the channels. Each cell was subjected to the same kinetic protocol before and after Ba²⁺ superfusion, which allowed us to superimpose the current traces (Figure 4A–C).

Display large version of this figure

Figure 4

Kinetics of Ba²⁺ block and unblock in WT KCNQ1 and mutants L273F and L273W. (A–C) After depolarization from −80mV to −20mV for 24s, 10mM Ba²⁺ was applied to oocytes expressing WT or mutant channels for 12s and then was washed out for 48s, while keeping the membrane at −20mV. (D) Fast (triangles) and slow (squares) time constants of Ba²⁺ block in WT KCNQ1 (black) and mutant L273F (blue) with their respective slow component amplitude (E) plotted as a function of voltage (A_s and A_f are the amplitudes of the slow and fast components, respectively). (F) Fast (open triangles) and slow (open squares) time constants of Ba²⁺ unblock in WT KCNQ1 (black) and mutant L273F (blue) with their respective slow component amplitudes (G) plotted as a function of voltage. Data shown are mean±SE (n=6–8).

We previously showed that in WT KCNQ1, Ba²⁺ block and unblock exhibit fast and slow kinetic components 34 (Figure 4A). When similar kinetic experiments are performed on the L273F mutant channels, well after they relax into slow inactivation, the rate of Ba²⁺ block and unblock is considerably slower than that of the WT (Figure 4B). The kinetics of the fast component of Ba²⁺ block is much slower in the mutant compared to WT, with a difference that increases with depolarization (Figure 4D, triangles). This difference becomes even more profound with depolarization when considering that the amplitude of the slow kinetic component increases in L273F and decreases in the WT (Figure 4E). Strikingly, both fast and slow components of Ba²⁺ unblock of L273F are far slower than in WT (Figure 4FG). Superimposition of the current traces shows that unblock and recovery of the current upon Ba²⁺ washout is so slow that it leads to a current deficit at the end of the protocol (Figure 4BA). This feature suggests that during slow inactivation, the selectivity filter assumes a conformation that traps Ba²⁺ and delays its egress from the deep pore. Compared to L273F mutant channels, the trapping of Ba²⁺ ions is not significant in WT KCNQ1 (Figure 4BA). The slow kinetics of Ba²⁺ unblock and the degree of current deficit depend on the time L273F channels spend in the slow inactivated state. Indeed, when Ba²⁺ is applied shortly after activation, kinetics of Ba²⁺ block and unblock is significantly faster than that obtained when Ba²⁺ is applied long after L273F channels relax into slow inactivation (Fig. 5).

Display large version of this figure

Figure 5

Dependence of Ba²⁺ trapping on the time mutant L273F spends in slow inactivation. Representative traces recorded without or with 10mM Ba²⁺ that was applied for 12s after 24s (A) or 0.9s (B) of depolarization to −20mV, followed by washout (48s). (C) Slow (open bars) and fast (solid bars) time constants of Ba²⁺ unblock after two different application times with (D) their respective slow (open bars) and fast (solid bars) amplitude components. Data shown are mean±SE (n=5; * p<0.005).

The link between Ba²⁺ trapping in the pore and the slow inactivation of mutant L273F is highly significant. Indeed, in mutant L273W, where no slow inactivation is observed, the kinetics of Ba²⁺ block and unblock is similar to that displayed by WT KCNQ1 (Figure 4C and Supplementary Material Fig. 2 ). Further correlation between slow inactivation and Ba²⁺ trapping is provided by the properties of other KCNQ1 pore mutants such as E295A in the turret, V310G in the pore helix, and D317A in the channel outer vestibule. We found that these pore mutants display a voltage-dependent slow inactivation which correlates with Ba²⁺ trapping (Fig. 6). Previous studies showed that substitution of the pore helix residue Val³¹⁰ with smaller amino acids like glycine or alanine produced voltage-dependent slow inactivation, with no change in K⁺ selectivity 44. We examined the kinetics of Ba²⁺ block and unblock in the inactivating mutant V310G. While modest slow inactivation and small current deficit are observed at 0mV (fractional inactivation=0.07±0.01; relative current deficit=0.14±0.02; n=6), substantial slow inactivation accompanied by a large current deficit are detected at +40mV (fractional inactivation=0.61±0.02; relative current deficit=0.32±0.01; n=6) (Figure 6A–C). The kinetics of Ba²⁺ block and unblock in the E295A mutant is very slow (Figure 6DE). At 0mV, only one exponential time constant describes the slow Ba²⁺ block (τ=4.63±0.48s, n=7), while a very slow unblock causes strong Ba²⁺ trapping (relative current deficit=0.48±0.04, n=7). Similarly to the turret (E295A) and pore helix (V310G) mutants, the outer vestibule mutant D317A generates a voltage-dependent slow inactivation which correlates with Ba²⁺ trapping. Interestingly, the D317A mutant slowly recovers from inactivation during a sustained depolarizing step (Figure 6F). Compared to WT, the kinetics of Ba²⁺ block of D317A is much slower, with the slow kinetic component being predominant. At +20mV, only one exponential time constant describes the slow Ba²⁺ unblock of D317A, leading to pronounced Ba²⁺ trapping (Figure 6F–I).

Display large version of this figure

Figure 6

Slow inactivation of KCNQ1 mutants causes Ba²⁺ trapping. Representative traces of mutant V310G at 0mV (A) or at +40mV (B) showing Ba²⁺ block (10mM, 12s) and unblock upon washout for 48s. (C) Current deficit upon Ba²⁺ washout (open bars) and fractional inactivation (solid bars) measured in mutant V310G at 0mV, +20mV, and +40mV. Current deficit=(Y-Z)/Y, where Z is the current measured after 48s Ba²⁺ washout and Y is the current at the same time point in the absence of Ba²⁺. (D) Representative traces of mutant E295A at 0mV showing Ba²⁺ block (12s) and unblock upon 48s washout. (E) Time constant of Ba²⁺ block in E295A channels at 0mV (left) and current deficit measured after 48s washout (right). (F) Representative traces of mutant D317A at +20mV showing Ba²⁺ block (12s) and unblock upon 48s washout. (G) Slow (open bars) and fast (solid bars) time constants of Ba²⁺ block (10s) in WT and D317A channels at +20mV, with their respective slow component amplitudes (H, I) Slow (open bars) and fast (solid bars) time constants of Ba²⁺ unblock in WT and mutant D317A. Data shown are mean±SE (n=5–7; * p<0.001).

In 2mM external K⁺, mutant L273F exhibits significant slower deactivation than WT. At a tail potential of −120mV, deactivation kinetics could be fitted by a two exponential function with τ_fast=54±10ms, τ_slow=308±65ms (relative slow component amplitude A_s/(A_f+A_s)=0.25±0.05) and τ_fast=104±9ms, τ_slow=527±85ms (A_s/(A_f+A_s)=0.44±0.03) for WT and L273F, respectively (n=6–8, p<0.01; Figure 7A–C). In contrast, the noninactivating mutant L273W displays similar deactivation kinetics to those of WT with τ_fast=61±11ms and τ_slow=304±45ms (A_s/A_f+A_s=0.26±0.04; n=6). In 50mM external K⁺, mutant L273F closes more slowly than it does in 2mM K⁺ and yet shows slower deactivation than WT with τ_fast=197±21ms, τ_slow=1593±521ms (A_s/(A_f+A_s)=0.51±0.08) (n=8, p<0.01, Figure 7D–F).

Display large version of this figure

Figure 7

Mutant L273F exhibits slower deactivation kinetics than WT. (A) Representative traces of WT (upper panel) and L273F channels (lower panel) recorded in 2mM external K⁺ with depolarizing steps from −70mV to +40mV in 10mV increments (holding potential −80mV; tail potentials −120mV). (B) Fast and slow deactivation time constants obtained from the biexponential fit of the deactivating tail currents in 2mM external K⁺ and (C) their corresponding relative slow component amplitude A_s/(A_s+A_f). (D) Representative traces of WT (upper panel) and L273F channels (lower panel) recorded in 50mM external K⁺. (E) Fast and slow deactivation time constants measured in 50mM external K⁺ and (F) their corresponding relative slow component amplitude. Data shown are mean±SE (n=6–8; * p<0.01).

Increasing external K⁺ concentration significantly accelerates the kinetics of slow inactivation of mutant L273F (Figure 8AB). In virtually 0K⁺, a single slow relaxation time constant is observed while at 50mM external K⁺ both fast and slow time constants are apparent, with the slow kinetic component being even much faster than that seen in the absence of K⁺ (Figure 8C). The fastest inactivation is observed at 50mM Rb⁺ ions (not shown). Interestingly, the speedup of the inactivation time course is coupled to an acceleration of the activation kinetics, a link specific for K⁺ ions. In high external K⁺ (50mM K⁺), the time to peak of the L273F channel activation is faster (at +30mV, t=99±8ms; n=5, p<0.01) than in high external Na⁺ (t=199±15ms, at 98mM Na⁺, n=5). The slowest channel activation is obtained with high external Li⁺ (50mM) whose time to peak value is t=240±23ms (n=6; p<0.01). The acceleration of activation and inactivation kinetics of L273F by external K⁺ is also found in other slowly inactivating KCNQ1 pore mutants (not shown). Notably, the kinetics of Ba²⁺ block and unblock in mutant L273F are significantly slower in high than in low external K⁺ (Figure 8D–F).

Display large version of this figure

Figure 8

Effect of external K⁺ on kinetics of activation, inactivation, and Ba²⁺ trapping in mutant L273F. (A) Representative I-V traces of L273F channels recorded in 50mM external K⁺ with depolarizing steps from −70mV to +30mV in 10mV increments (holding potential −80mV; tail potentials −120mV). (B) Representative normalized traces of L273F slow inactivation recorded from the same cell at +30mV in external solutions containing: 50mM Li⁺ and 48mM Na⁺; 2mM K⁺ and 96mM Na⁺; and 50mM K⁺ and 48mM Na⁺. (C) Time constants of slow inactivation displayed by L273F channels under the indicated external solutions. An additional fast time constant (solid bars) could be measured in solutions containing 50mM K⁺. (D) Slow (open bars) and fast (solid bars) time constants of Ba²⁺ block in L273F channels measured at +30mV in 2 and 50mM external K⁺. (E) Slow (open bars) and fast (solid bars) time constants of Ba²⁺ unblock in L273F channels measured at +30mV in 2 and 50mM external K⁺, with (F) their respective slow (open bars) and fast (solid bars) amplitude components. Data shown are mean±SE (n=6–8; * p<0.05, ** p<0.01).

In an attempt to understand how the L273F pore mutation leads to slow inactivation, we first constructed homology models in vacuum for the WT KCNQ1 and its pore mutants using the atomic coordinates of Kv1.2 as a template 48 (Fig. 9 and Supplementary Material Fig. 3 ). Comparison between the models of the WT KCNQ1 and the L273F mutant suggests that the mutation affects only the selectivity filter site s₁ and does not result in any conformational changes at sites s₂, s₃, and s₄. In this model, the phenylalanine is very close to the pore helix where it forms multiple short-distance hydrophobic (<3Å) and van der Waals (3–4Å) interactions with the main and side chains of Leu³⁰³ (Figure 9B and Supplementary Material data ). Consequently, the backbone flanking Leu³⁰³ moves together with Trp³⁰⁴. In both WT and mutant models, the Nɛ atom of Trp³⁰⁴ forms a hydrogen bond with the carboxyl of Asp³¹⁷, while its aromatic side chain also forms van der Waals interactions with the aromatic moiety of Tyr³¹⁵. As such, Trp³⁰⁴, D317, and Y315 move together and this movement introduces torsion in the backbone between positions 317 and 315. Consequently, the carbonyl oxygens of the four equatorial Tyr³¹⁵ residues in mutant L273F become closer to the axis of ion conduction, leading to constriction of the outer carbonyl ring of the selectivity filter. That is, the center-to-center distance between two facing carbonyl oxygens is 5.09Å versus 5.78Å in the L273F mutant versus WT, respectively.

Display large version of this figure

Figure 9

Structural models of WT and pore mutants of KCNQ1 channels. (A) Homology model of the pore domain of WT KCNQ1. For clarity, the frontal subunit was removed and ribbons of the other differently colored three subunits are shown from within the membrane. (B) Two diagonally facing subunits of mutant L273F (green) superposed on the homologous subunits of WT KCNQ1 (Corey Pauling, Kultin (CPK) colors: carbon, nitrogen, and oxygen atoms in light gray, blue, and red, respectively). The backbone of the selectivity filter and the indicated side chains are shown as sticks while part of S5 and the pore helix are shown as a Cα trace (in thinner lines). Other segments were removed for better viewing. Site s₁ corresponds to the place where a K⁺ ion is inferred to be trapped. Magnification of the backbone carbonyls of site s₁ is shown in black frame. (C) Superposition of mutant L273W (green) on the WT KCNQ1 (CPK) as in panel B. Dotted black lines correspond to distances of 2.6–2.9Å (hydrogen bonding between D317 and W304 and hydrophobic interactions between F273 and L303 in panel B). Dotted gray lines correspond to distances of ∼3.7Å (van der Waals interactions).

In the L273W mutant model, the side chain of Trp²⁷³ points downward and neither interacts with Leu³⁰³ nor changes the structure of the pore helix and thus does not lead to constriction of the outer carbonyl ring of the selectivity filter (Figure 9C). These observations are consistent with the inability of the L273W mutation to produce slow inactivation (Figure 4C). Interestingly, in all other in vacuum mutant models (D317A, E295A, and V310G), the structural changes emerging from the loss of the bond network of the replaced amino acids ultimately lead to constriction of the carbonyl ring at position 315 (see Supplementary Material Fig. 4 and Supplementary Data ).

It should be noted that the distances measured in vacuo across the selectivity filter reflect a static and limited picture of one given minimized structure. In an attempt to get better insight into the filter’s conformational changes that might accompany K⁺ conductance in WT KCNQ1 and the various pore mutants, we ran molecular dynamics simulations using the aforementioned models in the presence of water and potassium ions (Fig. 10 and Table 1). We found that the distances obtained in the dynamics simulations are within the range of those displayed by the in vacuum models. However, the simulations show that the mean distance between opposite carbonyls at position Y315 in WT and the inactivating mutants are not significantly different (Table 1). Notably, this mean distance does not reflect the extent of the breathing motions of the selectivity filter. In this regard, the maximal distance between facing carbonyl oxygens is more meaningful. Indeed, the simulations show that in all slow inactivating mutants this maximal distance is smaller than that of the WT and the noninactivating mutant L273W (Table 1). The distribution of the distances between the carbonyls of Tyr³¹⁵ is shifted in the L273F mutant relative to the WT (Figure 10D). The corresponding distribution for the L273W mutant does not change significantly compared to the WT (Figure 10E). Notably, the range of the fluctuations in the carbonyl distance of Tyr³¹⁵ in the slow inactivating mutants L273F, E295A, V310G, and D317A decreases compared to the WT and the L273W mutant (Table 1). This suggests that the L273F, E295A, V310G, and D317A mutations also lead to a substantial decrease in the flexibility of this region of the selectivity filter.

Display large version of this figure

Figure 10

Molecular dynamics simulation with a primary hydration shell and three K⁺ ions. (A–C) Distances between the carbonyls of Y315 in two opposite facing subunits as a function of time for (A) WT KCNQ1, (B) the L273F, and (C) L273W mutants. (D and E) Distribution of the distances between the carbonyls of Y315 in two opposite facing subunits of the WT KCNQ1 and the (D) L273F and (E) L273W mutants.

Quantitative indication concerning the effect of the slow inactivating mutants on the flexibility of the selectivity filter is obtained through the analysis of the third and fourth moments about the mean of the distribution of the distances between the carbonyls of Tyr³¹⁵ (Table 1). While the first moment, the mean, and the second moment, the variance, characterize the location and the variability of a distribution, respectively, the third and fourth moments characterize the shape of a distribution. The third moment, the skewness, provides an indication about the departure of a distribution from symmetry. Compared with a normal distribution that has a skewness of 0, all distributions were only slightly positively skewed. The extent to which both the WT and the L273W mutant were skewed was, however, higher than that of the slow inactivating mutants (L273F, E295A, V310G, and D317A). For both the WT and the noninactivating mutant L273W, the skewness was 0.69, while for the L273F mutant, it was 0.27. This difference is significant as the standard error of skewness is estimated as 0.05 according to where n=3000 is the number of structures included in the analysis 49. This is a quantitative indication that the distribution of the distances between the carbonyls of Tyr³¹⁵ in the WT and the L273W mutant has an asymmetric tail extending further to larger distances than that of the slow inactivating L273F mutant. The fourth moment, the kurtosis, is a measure of the peakedness of the distribution and of the heaviness of its tails 50. Since a normal distribution has kurtosis of three, the excess kurtosis relative to normal distribution is used and obtained by subtracting three from the value of the fourth moment around the mean. The higher the kurtosis of a distribution, the more distinct its peak close to the mean, the more rapid its decline, and the heavier its tails. As the variance for the distributions is roughly the same, we have proceeded to compare the excess kurtosis of these distributions. At constant variance, a higher kurtosis is an indication of a larger number of cases that deviate further away from the mean. Comparison of the excess kurtosis shows that while for the WT and the noninactivating mutant L273W, the value is approximately the same, and equal to 1.32 and 1.36, respectively, the value for the slow inactivating mutants (L273F, E295A, V310G, and D317A) is smaller (Table 1). For example, excess kurtosis equals to 0.53 for mutant L273F. In view of the large number of structures (n=3000) that are included in the analysis, the standard error of kurtosis can be estimated as 0.09 according to 49, indicating that this difference is significant. As a numerical measure, the excess kurtosis further confirms that the extent to which the distance between the carbonyls of Tyr³¹⁵ can increase further away from the mean is larger for the WT and the noninactivating mutant L273W than for the slow inactivating mutants L273F, E295A, V310G, and D317A.

In all, the dynamics simulations indicate that compared to the WT and the noninactivating mutant L273W, the slow inactivating mutants exhibit both a reduced maximal distance between the facing carbonyls of Y315 and smaller values of excess kurtosis. This implies that in the slow inactivating pore mutants, the upper filter has a reduced capacity to fluctuate, namely, an increased rigidity.

Inactivation of WT KCNQ1 is invisible from macroscopic inspection but a careful analysis indicates that it is an intrinsically fast and voltage-independent process 30,31,32. Interestingly, KCNQ1 pore mutants generate an additional voltage-dependent slow inactivation 42,43,44. Here we show that these two types of inactivation are discrete processes, displaying distinct kinetics and pore properties. The pore mutant L273F exhibits a fast inactivation similar to the WT, with fast relaxation and recovery time courses. In addition, this channel mutant displays a voltage-dependent slow inactivation, with very slow relaxation and recovery kinetics. Further distinctions between these two inactivation types is provided by their differential impact on kinetics of Ba²⁺ block and unblock, which revealed that Ba²⁺ trapping deep in the pore arises primarily from the slow inactivation.

Inactivation of KCNQ1 channels does not share the characteristic features of N- and C-type inactivation. The lack of an N-terminal ball structure within KCNQ1 and the persistence of both types of inactivation in an N-terminally truncated short isoform of mutant L273F (not shown) indicate that KCNQ1 inactivation differs from N-type inactivation. Moreover, the simulations based on the kinetic model that we elaborated here, clearly show that the gating behavior of mutant L273F gives rise to an additional inactivation state (I₂), which reflects the slow inactivation. The latter is accelerated by elevated concentrations of external K⁺, unlike C-type inactivation of Shaker K⁺ channels which is inhibited by increased external K⁺ concentration 15,23.

Ba²⁺ has a large enthalpy of hydration (ΔH=−328 kcal/mol) 51 and, like K⁺, must be dehydrated before reaching its deep binding site in the selectivity filter and must be resolvated upon its egress from the pore to the external solution. Our experimental results suggest that during slow inactivation the upper part of the selectivity filter obstructs the path of Ba²⁺ to and from its deep site. This interpretation is supported by the findings that during slow inactivation, Ba²⁺ block and unblock are markedly delayed despite the steady-state affinity and the electric distance for Ba²⁺ block are similar in WT and mutant L273F. It would be interesting to investigate the relationship between inactivation and Ba²⁺ trapping at the single-channel level. However, this is hampered by the very small unitary conductance of KCNQ1 52,53,54.

Several lines of evidence point to a tight link between slowdown of Ba²⁺ entry, Ba²⁺ trapping, and slow inactivation:

Trapping of Ba²⁺ ions is minimal in WT KCNQ1 channels, raising the question of what might be the physical determinants of fast inactivation. The inactivation of WT KCNQ1 was suggested to relate to fast flickering of the open channel, which may reflect instability of a fast voltage-independent pore gate 35,55. The fast inactivation of WT KCNQ1 does not lead to a large current decline, unlike the conductance fall-down produced by the slow inactivation of the L273F pore mutant described here. The fast voltage-independent flicker possibly reflects small-scale fluctuating motions of the selectivity filter that are not coupled to the sensor movement.

The L273F mutant is a typical example of a group of human inherited mutations which produce a voltage-dependent slow inactivation and are responsible for the long QT syndrome. Indeed, we recently showed that LQT mutations located in the proximal C terminus of KCNQ1 (R366W, A371T, S373P, and W392R) impair calmodulin binding and alter channel gating by generating a voltage-dependent slow inactivation 56. A similar slow inactivation gating behavior is recurrently found in mutated residues of many regions of KCNQ1. Notably, mutations in the voltage sensor S4 (L233W and Q244W) produce a voltage-dependent slow inactivation (Supplementary Material Fig. 5 ), which suggests that the movement of the voltage sensor promotes the slow inactivation. In Shaker-related Kv channels, it appears that slow C-type inactivation is also coupled to voltage sensor movement 57,58,59,60,61,62.

No conformational changes are observed at filter sites s₂, s₃, and s₄ in the model of L273F, in line with our results showing that the electrical distance and affinity of Ba²⁺ to its innermost filter site s₄ are similar in mutant L273F and the WT. In contrast, the in vacuum model predicts conformational changes that converge to the upper part of the selectivity filter. We suggest that a decreased flexibility at this part of the filter, as revealed by the dynamics simulations, leads to tighter coordination of a K⁺ ion in s₁. This, in turn, causes hindrance of the dehydration-resolvation transition, impairing thereby high throughput conduction. Our interpretations are in line with the acceleration of slow inactivation and the slower kinetics of Ba²⁺ block and unblock observed upon elevation of external K⁺.

The KCNQ1 slow inactivation is largely different from the foot-in-the-door mechanism described for C-type inactivation where the pore is presumed to adopt a nonconductive conformation when permeant ions vacate the selectivity filter 19,20,21. Interestingly, other channel types also exhibit inactivation properties that diverge from the properties originally described for C-type in Shaker. In rapidly inactivating Kv4 channels, elevation of extracellular K⁺ concentration promotes inactivation and inhibits recovery of the channels from inactivation 63,64,65. Similarly, Kv2.1 inactivation is accelerated at elevated concentrations of extracellular K⁺ or TEA⁺66. In contrast, for C-type inactivation, increase in extracellular K⁺ is considered to be slow depletion of K⁺ ions from the selectivity filter and thus to prevent inactivation. However, a coherent picture of C-type inactivation has not emerged yet. The x-ray crystal structure of KcsA determined at low K⁺ concentration was suggested to reflect a nonconducting filter conformation 4,67. Recent molecular dynamics simulations of KcsA indicate that under low K⁺ occupancy, the selectivity filter changes its conformation to a nonconducting state of broken fourfold symmetry with widening of the filter, loss of K⁺ coordination at s₂, and stronger K⁺ coordination at s₃68. Along this view, it is assumed that, in KcsA, the hydrogen bonding between Tyr⁷⁸ of the GYG motif or Asp⁸⁰ of the outer mouth and the aromatic cuff residues (Trp⁶⁷ and Trp⁶⁸) is very important for stabilization of the conducting conformation of the selectivity filter 19,68. On the other hand, combined structural and functional studies have recently challenged this view by showing that in KcsA the hydrogen bond between Asp⁸⁰ and Trp⁶⁷ and the carboxy-carboxylate interaction between Asp⁸⁰ and Glu⁷¹ destabilize the conducting conformation of the KcsA selectivity filter 14. Disrupting these interactions stabilizes a noninactivating, high-open probability channel state. According to this view, the low K⁺ crystal structure of KcsA could not represent the C-type inactivated state of the channel 14. Our results show that disruption of the hydrogen bonding between Trp³⁰⁴ and Asp³¹⁷ in KCNQ1 (mutant D317A) destabilizes a conducting filter conformation and produces slow inactivation associated with Ba²⁺ trapping (Figure 6F–I).

Molecular modeling and dynamics simulations indicate that even minimal changes can already be effective by reducing the flexibility of the upper part of the selectivity filter. The ability of the WT and the noninactivating mutant L273W to extend to wider distances may facilitate the entry to and exit of Ba²⁺ ions from the pore, thereby preventing Ba²⁺ trapping. The fluctuations rather than the average distance between the carbonyls of Tyr³¹⁵ appear to account for the difference between WT and the slow inactivating pore mutants. In other words, a decrease in breathing motions of the filter leads to an increase in the time K⁺ ions reside in the filter. Our experimental data are consistent with this notion of altered pore occupancy. We showed that the inactivating mutant L273F exhibits slower deactivation than the WT and that high external K⁺ further slows down channel closure (Fig. 7). This slower deactivation reflects longer residency time of K⁺ in the pore of mutant L273F (via O₁), which ultimately feeds back to the slow inactivation state I₂. This feature may also explain why increase in extracellular K⁺ accelerates inactivation. However, it is also possible that in the outer channel pore, there is a K⁺-selective binding site that modulates inactivation gating, thereby affecting the conformation of the selectivity filter. In all, decreased flexibility at filter site s₁ plausibly underlies the mechanism of slow inactivation in KCNQ1 pore mutants.